It's problematic because of the demographics of the areas and of the people who are predicted to commit crimes, of course. If the system doesn't spit out a perf

It's problematic because of the demographics of the areas and of the people who are predicted to commit crimes, of course. If the system doesn't spit out a perfectly egalitarian, proportional representation of the population down to an infinite level of granularity, it must be wrong; just like all the other machine learning systems that have been revealed to have incredible bias by similarly pointing out things that agree with observable, measured reality.